Database Studio — AI Copilot, PostgreSQL, MySQL, SQLite & Redis Client

AI Security Review

Evidence context: threat category unknown malware; evidence quality moderate.

The stated purpose is a database studio with an AI Copilot, database browsing and editing, .env auto-connect, and OpenRouter access. Reading workspace files for connection settings fits that purpose, while database editing can justify write actions against configured databases. The listing declares no extension permissions or host permissions, so the available metadata does not establish broad filesystem or process rights. The main code findings are OBFUSCATION-UNICODE_HEAVY-extension/out/extension.js-1 at extension/out/extension.js:1:0 and OBFUSCATION-HEX_STRING_HEAVY-extension/out/extension.js-10 at extension/out/extension.js:10:0. Those findings describe concealment patterns in the compiled extension entry point, rather than a clear database, shell, or child-process action. The stated database function makes workspace and database access reasonable, while the obfuscation leaves the exact scope of that access unclear.

The credential question is narrower. The product description promises “3-second .env auto-connect,” so reading .env files is within the advertised connection workflow. No finding title identifies .ssh, cloud credential directories, VS Code secret storage, browser cookies, or other unrelated secret stores. The available findings also contain no secret finding and no malware signature. A .env reader still handles database passwords when those values are present, so the code path at extension/out/extension.js:1:0 deserves direct inspection for collection and transmission behavior.

The network list includes openrouter.ai, which matches the advertised AI query feature, and sidorares.github.io, which is consistent with a JavaScript database driver. The domains ciromaciel.click, by-ciromaciel.click, ae.ci, he.no, ii.no, and qd.ca have no stated role in the product description. Their presence alone does not prove exfiltration because no network finding connects them to workspace contents or credentials. It does create an unresolved question about telemetry, update checks, redirects, or bundled dependency traffic.

The strongest counterargument is that extension/out/extension.js:1:0 is compiled output. Minified or bundled JavaScript can trigger Unicode-heavy and hexadecimal-heavy rules without malicious intent, and the absence of malware signatures supports that explanation. That argument does not settle the issue because both high-severity obfuscation titles target the extension entry point, while several unrelated external domains remain listed and the product handles .env database credentials. Runtime tracing and inspection of the unpacked source or source map should determine whether those domains receive query text, workspace data, or secrets.

Key Reasons

• OBFUSCATION-UNICODE_HEAVY-extension/out/extension.js-1 and OBFUSCATION-HEX_STRING_HEAVY-extension/out/extension.js-10 target the compiled extension entry point.
• The advertised .env auto-connect feature can handle database passwords, even though no finding names SSH keys, cloud credentials, or secret storage.
• openrouter.ai fits the AI feature, while ciromaciel.click, by-ciromaciel.click, ae.ci, he.no, ii.no, and qd.ca have no stated product role.
• No malware signature or network finding establishes malicious collection or exfiltration.

False Positive Considerations

• Compiled output at extension/out/extension.js:1:0 can trigger Unicode-heavy obfuscation rules.
• Compiled output at extension/out/extension.js:10:0 can trigger hexadecimal-heavy obfuscation rules.
• openrouter.ai and sidorares.github.io fit the advertised AI and database-driver functions.
• No malware signature or secret finding connects the listed domains to stolen data.

Reviewed 2026-09-29; recommended action: runtime analysis; model confidence 76%.